Area-based face curve characteristic analysis to recognize Multimodal 2D/3D monozygotic twins using Simpson’s rule and Machine Learning

Abstract

Recent advances in face recognition have achieved high accuracy in identifying individuals. However, distinguishing identical twins remains challenging due to their substantial facial similarity. Human vision and collective intelligence suggest that the lower face margin curve is the most distinctive region for differentiating twins. Hence, this proposed technique measures and compares the face curve characteristics of the identical twins by calculating the area of the face curve using Simpson’s rules from values of the ordinates about the face’s vertical axis along the nose point. To more accurately identify and analyze the facial differences and compare the twin faces, the resulting area-based score is then used as input to various machine learning algorithms such as Extreme gradient boosting (XGBoost), Adaptive Boosting (AdaBoost) classifiers, Random Forest (RF) classifiers, Light Gradient Boosting Model(LGBM), and Extra Tree Classifier(ETC) classifiers, etc. The datasets ND-TWINS and 3D TEC produce encouraging classification rates of 94%, and 86%. In this paper, we discuss the impact of Simpson’s rule on categorical data and demonstrate its effects on AI and ML application scenarios.